與「Kenton Lee」相符的使用者個人學術檔案
 | Kenton LeeGoogle DeepMind 在 google.com 的電子郵件地址已通過驗證 被引用 138917 次 |
End-to-end neural coreference resolution
We introduce the first end-to-end coreference resolution model and show that it significantly
outperforms all previous work without using a syntactic parser or hand-engineered mention …
outperforms all previous work without using a syntactic parser or hand-engineered mention …
Bert: Pre-training of deep bidirectional transformers for language understanding
We introduce a new language representation model called BERT, which stands for Bidirectional
Encoder Representations from Transformers. Unlike recent language representation …
Encoder Representations from Transformers. Unlike recent language representation …
Natural questions: a benchmark for question answering research
…, D Epstein, I Polosukhin, J Devlin, K Lee… - Transactions of the …, 2019 - direct.mit.edu
We present the Natural Questions corpus, a question answering data set. Questions consist
of real anonymized, aggregated queries issued to the Google search engine. An annotator …
of real anonymized, aggregated queries issued to the Google search engine. An annotator …
Higher-order coreference resolution with coarse-to-fine inference
We introduce a fully differentiable approximation to higher-order inference for coreference
resolution. Our approach uses the antecedent distribution from a span-ranking architecture as …
resolution. Our approach uses the antecedent distribution from a span-ranking architecture as …
Retrieval augmented language model pre-training
Abstract Language model pre-training has been shown to capture a surprising amount of
world knowledge, crucial for NLP tasks such as question answering. However, this knowledge …
world knowledge, crucial for NLP tasks such as question answering. However, this knowledge …
Pix2struct: Screenshot parsing as pretraining for visual language understanding
Visually-situated language is ubiquitous—sources range from textbooks with diagrams to web
pages with images and tables, to mobile apps with buttons and forms. Perhaps due to this …
pages with images and tables, to mobile apps with buttons and forms. Perhaps due to this …
Legibility and predictability of robot motion
A key requirement for seamless human-robot collaboration is for the robot to make its intentions
clear to its human collaborator. A collaborative robot's motion must be legible, or intent-…
clear to its human collaborator. A collaborative robot's motion must be legible, or intent-…
Latent retrieval for weakly supervised open domain question answering
Recent work on open domain question answering (QA) assumes strong supervision of the
supporting evidence and/or assumes a blackbox information retrieval (IR) system to retrieve …
supporting evidence and/or assumes a blackbox information retrieval (IR) system to retrieve …
BoolQ: Exploring the surprising difficulty of natural yes/no questions
In this paper we study yes/no questions that are naturally occurring --- meaning that they are
generated in unprompted and unconstrained settings. We build a reading comprehension …
generated in unprompted and unconstrained settings. We build a reading comprehension …
Well-read students learn better: On the importance of pre-training compact models
Recent developments in natural language representations have been accompanied by
large and expensive models that leverage vast amounts of general-domain text through self-…
large and expensive models that leverage vast amounts of general-domain text through self-…